BackgroundIn recent years, research in artificial neural networks has resurged, now under the deep-learning umbrella, and grown extremely popular. Recently reported success of DL techniques in crowd-sourced QSAR and predictive toxicology competitions has showcased these methods as powerful tools in drug-discovery and toxicology research. The aim of this work was dual, first large number of hyper-parameter configurations were explored to investigate how they affect the performance of DNNs and could act as starting points when tuning DNNs and second their performance was compared to popular methods widely employed in the field of cheminformatics namely Naïve Bayes, k-nearest neighbor, random forest and support vector machines. Moreover, robustness of machine learning methods to different levels of artificially introduced noise was assessed. The open-source Caffe deep-learning framework and modern NVidia GPU units were utilized to carry out this study, allowing large number of DNN configurations to be explored.ResultsWe show that feed-forward deep neural networks are capable of achieving strong classification performance and outperform shallow methods across diverse activity classes when optimized. Hyper-parameters that were found to play critical role are the activation function, dropout regularization, number hidden layers and number of neurons. When compared to the rest methods, tuned DNNs were found to statistically outperform, with p value <0.01 based on Wilcoxon statistical test. DNN achieved on average MCC units of 0.149 higher than NB, 0.092 than kNN, 0.052 than SVM with linear kernel, 0.021 than RF and finally 0.009 higher than SVM with radial basis function kernel. When exploring robustness to noise, non-linear methods were found to perform well when dealing with low levels of noise, lower than or equal to 20%, however when dealing with higher levels of noise, higher than 30%, the Naïve Bayes method was found to perform well and even outperform at the highest level of noise 50% more sophisticated methods across several datasets.Electronic supplementary materialThe online version of this article (doi:10.1186/s13321-017-0226-y) contains supplementary material, which is available to authorized users.

No abstract

A quadrotor is a rotorcraft capable of hover, forward flight, and VTOL and is emerging as a fundamental research and application platform at present with flexibility, adaptability, and ease of construction. Since a quadrotor is basically considered an unstable system with the characteristics of dynamics such as being intensively nonlinear, multivariable, strongly coupled, and underactuated, a precise and practical model is critical to control the vehicle which seems to be simple to operate. As a rotorcraft, the dynamics of a quadrotor is mainly dominated by the complicated aerodynamic effects of the rotors. This paper gives a tutorial of the platform configuration, methodology of modeling, comprehensive nonlinear model, the aerodynamic effects, and model identification for a quadrotor.

No abstract

A regional ensemble prediction system (REPS) with the limited-area version of the Canadian Global Environmental Multiscale (GEM) model at 15-km horizontal resolution is developed and tested. The total energy norm singular vectors (SVs) targeted over northeastern North America are used for initial and boundary perturbations. Two SV perturbation strategies are tested: dry SVs with dry simplified physics and moist SVs with simplified physics, including stratiform condensation and convective precipitation as well as dry processes. Model physics uncertainties are partly accounted for by stochastically perturbing two parameters: the threshold vertical velocity in the trigger function of the Kain-Fritsch deep convection scheme, and the threshold humidity in the Sundqvist explicit scheme. The perturbations are obtained from firstorder Markov processes. Short-range ensemble forecasts in summer with 16 members are performed for five different experiments. The experiments employ different perturbation and piloting strategies, and two different surface schemes. Verification focuses on quantitative precipitation forecasts and is done using a range of probabilistic measures. Results indicate that using moist SVs instead of dry SVs has a stronger impact on precipitation than on dynamical fields. Forecast skill for precipitation is greatly influenced by the dominant synoptic weather systems. For stratiform precipitation caused by strong baroclinic systems, the forecast skill is improved in the moist SV experiments relative to the dry SV experiments. For convective precipitation rates in the range 15-50 mm (24 h) Ϫ1 produced by weak synoptic baroclinic systems, all experiments exhibit noticeably poorer forecast skills. Skill improvements due to the Interactions between Soil, Biosphere, and Atmosphere (ISBA) surface scheme and stochastic perturbations are also observed.